Download A Comparison of Two Word-Recognition Tasks in Multitalker Babble

J Am Acad Audiol 19:548–556 (2008)
A Comparison of Two Word-Recognition Tasks in
Multitalker Babble: Speech Recognition in Noise Test
(SPRINT) and Words-in-Noise Test (WIN)
DOI: 10.3766/jaaa.19.7.4
Richard H. Wilson*
Wendy B. Cates*
Abstract
Background: The Speech Recognition in Noise Test (SPRINT) is a word-recognition instrument that
presents the 200 Northwestern University Auditory Test No. 6 (NU-6) words binaurally at 50 dB HL in a
multitalker babble at a 9 dB signal-to-noise ratio (S/N) (Cord et al, 1992). The SPRINT was developed
by and used by the Army as a more valid predictor of communication abilities (than pure-tone
thresholds or word-recognition in quiet) for issues involving fitness for duty from a hearing perspective
of Army personnel. The Words-in-Noise test (WIN) is a slightly different word-recognition task in a fixed
level multitalker babble with 10 NU-6 words presented at each of 7 S/N from 24 to 0 dB S/N in 4 dB
decrements (Wilson, 2003; Wilson and McArdle, 2007). For the two instruments, both the babble and
the speakers of the words are different. The SPRINT uses all 200 NU-6 words, whereas the WIN uses
a maximum of 70 words.
Purpose: The purpose was to compare recognition performances by 24 young listeners with normal
hearing and 48 older listeners with sensorineural hearing on the SPRINT and WIN protocols.
Research Design: A quasi-experimental, mixed model design was used.
Study Sample: The 24 young listeners with normal hearing (19 to 29 years, mean 5 23.3 years) were
from the local university and had normal hearing (#20 dB HL; American National Standards Institute,
2004) at the 250–8000 Hz octave intervals. The 48 older listeners with sensorineural hearing loss (60
to 82 years, mean 5 69.9 years) had the following inclusion criteria: (1) a threshold at 500 Hz between
15 and 30 dB HL, (2) a threshold at 1000 Hz between 20 and 40 dB HL, (3) a three-frequency puretone average (500, 1000, and 2000 Hz) of #40 dB HL, (4) word-recognition scores in quiet $40%, and
(5) no history of middle ear or retrocochlear pathology as determined by an audiologic evaluation.
Data Collection and Analysis: The speech materials were presented bilaterally in the following order:
(1) the SPRINT at 50 dB HL, (2) two half lists of NU-6 words in quiet at 60 dB HL and 80 dB HL, and (3)
the two 35-word lists of the WIN materials with the multitalker babble fixed at 60 dB HL. Data collection
occurred during a 40–60 minute session. Recognition performances on each stimulus word were
analyzed.
Results: The listeners with normal hearing obtained 92.5% correct on the SPRINT with a 50% point on
the WIN of 2.7 dB S/N. The listeners with hearing loss obtained 65.3% correct on the SPRINT and a
WIN 50% point at 12.0 dB S/N. The SPRINT and WIN were significantly correlated (r 5 20.81, p ,
.01), indicating that the SPRINT had good concurrent validity. The high-frequency, pure-tone average
(1000, 2000, 4000 Hz) had higher correlations with the SPRINT, WIN, and NU-6 in quiet than did the
traditional three-frequency pure-tone average (500, 1000, 2000 Hz).
Conclusions: Graphically and numerically the SPRINT and WIN were highly related, which is indicative
of good concurrent validity of the SPRINT.
Key Words: Auditory perception, hearing loss, speech perception, word recognition in multitalker babble
*VA Medical Center, Mountain Home, TN, and Departments of Surgery and Communicative Disorders, East Tennessee State University,
Johnson City, TN
Richard H. Wilson, Ph.D, VA Medical Center, Audiology (126), Mountain Home, TN 37684; Phone: 423-979-3561; Fax: 423-9793403; E-mail: [email protected]
This work was presented at AudiologyNOW! 2008 in Charlotte, NC.
548
Word Recognition in Noise Tests—SPRINT and WIN/Wilson and Cates
Abbreviations: HFPTA 5 high-frequency, pure-tone average; MOS 5 Military Occupational Specialty;
MMRB 5 Medical MOS Retention Board; NU-6 5 Northwestern University Auditory Test No. 6; S/
N 5 signal-to-noise ratio; SPRINT 5 Speech Recognition in Noise Test; VA 5 Veterans Affairs;
WIN 5 Words-in-Noise Test
Sumario
Antecedentes: La Prueba de Reconocimiento del Lenguaje en Ruido (SPRINT) es un instrumento de
reconocimiento de palabras que presenta binauralmente las 200 palabras de la Prueba Auditiva No. 6
de la Universidad Northwestern a 50 dB HL en balbuceo de hablantes múltiples, a una tasa señal-ruido
(S/N) de 9 dB (Cord y col., 1992). El SPRINT fue desarrollado y usado por el Ejército como un
elemento con mayor validez de predicción de las habilidades de comunicación (que los umbrales de
tonos puros o el reconocimiento de palabras en silencio) con asuntos relacionados con la aptitud para
el servicio militar desde la perspectiva auditiva del personal del ejército. La Prueba de Palabras en
Ruido (WIN) es una tarea ligeramente diferente de reconocimiento de palabras a un nivel fijo de
balbuceo de hablantes múltiples con 10 palabras NU-6 presentadas a cada uno de 7 S/N desde 24 a
0 dB S/N en decrementos de 4 dB (Wilson, 2003; Wilson y McArdle, 2007). Para los dos instrumentos,
tanto el balbuceo como los hablantes de las palabras son diferentes. El SPRINT usa todas las 200
palabras del NU-6, mientras que el WIN usa un máximo de 70 palabras.
Propósito: El propósito fue comparar el desempeño en reconocimiento de 24 sujetos con audición
normal y 48 sujetos mayores con alteración sensorineural, en relación con los protocolos SPRINT y
WIN.
Diseño de Investigación: Se utilizó un diseño mixto de modelo, cuasi- experimental.
Muestra del Estudio: Los 24 sujetos jóvenes con audición normal (19 a 29 años, media 5 23.3 años)
pertenecı́an a la universidad local y tenı́an umbrales normales (# 20 dB HL; Instituto Nacional
Americano de Estándares, 2004) en los intervalos de octava de 250–8000 Hz. Los 48 sujetos mayores
con hipoacusia sensorineural (60 a 82 años, media 5 69.9 años) tenı́an los siguientes criterios de
inclusión: (1) un umbral en 500 Hz entre 15 y 30 dB HL; (2) un umbral a 1000 Hz entre 20 y 40 db HL,
(3) un promedio tonal puro en tres frecuencias (500, 1000 y 2000 Hz) de #40 dB HL, (4) puntajes de
reconocimiento de palabras en silencio $40%, y (5) sin historia de patologı́a en el oı́do medio o
retrococlear, determinado por una evaluación audiológica.
Recolección y Análisis de Datos: Los materiales de lenguaje fueron presentados bilateralmente en el
siguiente orden: (1) el SPRINT a 50 dB HL, (2) dos medias listas de las palabras NU-6 en silencio a
60 dB HL y 80 dB HL, y (3) las dos listas de 35 palabras de los materiales del WIN con el balbuceo de
hablantes múltiples a un nivel fijo de 60 dB HL. La recolección de datos ocurrió durante una sesión de
40–60 minutos. Se analizó el desempeño en el reconocimiento en cada palabra estı́mulo.
Resultados: Los sujetos con audición normal obtuvieron un 92.5 % de respuestas correctas en el
SPRINT con un punto 50% en el WIN de S/N de 2.7 dB. Los sujetos con hipoacusia obtuvieron un
65.3% de respuestas correcta en el SPRINT y un punto 50% en el WIN a una S/N de 12.0 dB HL. El
SPRINT y el WIN correlacionaron significativamente (r 5 20.81, p , 0.01), indicando que el SPRINT
tuvo una buena validez concurrente. El promedio tonal a alta frecuencia (1000, 2000, 4000 Hz) tuvo
correlaciones mayores con el SPRINT, WIN y NU-6 en silencio que lo que tuvo el promedio tonal puro
tradicional en tres frecuencias (500, 1000 y 2000 Hz).
Conclusiones: Gráfica y numéricamente, el SPRINT y el WIN estuvieron altamente relacionados, lo
que es indicativo de una buena validez concurrente del SPRINT.
Palabras Clave: Percepción auditiva, pérdida auditiva, percepción del lenguaje, reconocimiento del
lenguaje en balbuceo de hablantes múltiples
Abreviaturas: HFPTA 5 promedio tonal puro de altas frecuencias; MOS 5 Especialidad Ocupacional
Militar; MMRB 5 Consejo de Retención del MOS Médico; NU-6 5 Prueba Auditiva No. 6 de la
Universidad Northwestern; S/N 5 tasa señal-ruido; SPRINT 5 Prueba de Reconocimiento del Lenguaje
en Ruido; VA 5 Asuntos de Veteranos; WIN 5 Prueba de Palabras en Ruido
T
he Army uses a physical profile system known
as PUHLES to categorize the general physical
condition, upper extremities, hearing, lower
extremities, vision (eye), and mental status of soldiers.
Within each category a four-point scale is incorporated
with 1 being the best rating and 4 being the worst
rating. Accordingly, the Army classifies hearing loss
by pure-tone thresholds using H-1 to H-4 profiles
549
Journal of the American Academy of Audiology/Volume 19, Number 7, 2008
Table 1. Decibel Hearing Levels Used to Define the H-1
and H-2 Profiles
Frequency in Hz
Profile
H-1 (both ears)
H-2 (both ears)
or
H-2 (better ear)
500
30
35
1000
30
35
2000
30
35
30
25
25
3 Freq. PTA
25
30
4000
45
55
35
Note: For the first two profiles, the unaided three-frequency puretone averages (500, 1000, and 2000 Hz) for each ear are the first
determiner with no single threshold being greater than the limits
indicated. For the third profile the better ear has to have the
indicated minimum thresholds and the poorer ear ‘‘may be deaf’’
(Department of the Army, 2007, table 7-1).
(Department of the Army, 2007). For the H-1 and H-2
profiles, hearing loss (unaided) in either ear can have
no pure-tone threshold greater than the limits indicated in Table 1 for the respective categories. If a
threshold exceeds the limits at one frequency, then the
next higher profile is applied. The H-3 profile, which
involves any pure-tone threshold that exceeds the
limits set for the H-2 profile, requires an audiological
evaluation and for fitness of duty may involve a
Medical Military Occupational Specialty (MOS) Retention Board, or MMRB, who determines if the soldier
should be (1) retained in his or her current MOS, (2)
reclassified into another MOS, or (3) discharged from
the Army. The H-4 profile indicates that hearing, even
with a hearing aid(s), is sufficiently impaired to
preclude continued service in the Army.
Despite its reliance on pure-tone thresholds to
profile hearing, both Army audiologists and members
of MMRB realized that neither pure-tone thresholds
nor speech-recognition abilities assessed in quiet were
accurate indicators of the ability of a soldier to
communicate in realistic environments involving background noise. Consequently, the Speech Recognition in
Noise Test (SPRINT) was developed as a more valid
predictor of communication abilities than were puretone thresholds or word recognition in quiet, that is,
fitness for duty from a hearing perspective (Cord et al,
1992). The SPRINT scores, in conjunction with other
medical, occupational, and administrative considerations, are used by the MMRB to make the most
appropriate determination regarding retention, reclassification, or discharge of the soldier. Administration of
the SPRINT is required by Army regulations for
soldiers having H-3 profiles whose medical records
are subject to review by an MMRB. The Army
administers thousands of SPRINTs annually. Additionally, in the current era of Iraqi Freedom and
Enduring Freedom, many private audiologists and VA
(Veterans Affairs) audiologists are being requested to
administer the SPRINT on military personnel being
activated and inactivated.
550
The SPRINT involves presentation of monosyllabic
words in background noise simultaneously to both
ears. Specifically, the 200 monosyllabic words that
constitute the Northwestern University Auditory Test.
No. 6 (NU-6; Tillman and Carhart, 1966) are presented
in a multitalker babble at one (9 dB) signal-to-noise
ratio (S/N). In keeping with the purpose for which the
test is administered, the large number of words is
included to achieve highly stable scores, following the
binomial model (Thornton and Raffin, 1978). The 9 dB
S/N was intended to produce 95–100% performances in
young adults with normal hearing, but widely varying
performances among soldiers having the H-3 profiles.
During development, the SPRINT was administered to
319 soldiers with H-3 profiles (Cord et al, 1992). Mean
recognition performance was 81.5% correct (163 of 200
words) with a range from 32.5% (65 words correct) to
98% (196 words correct). Test-retest recognition performances, which were established on 30 soldiers, were
82.3 and 83.4%, respectively, and were highly correlated (r 5 0.93, p , .01).
Despite being developed for the specific purpose
described above, use of the SPRINT has been proposed
for other applications within the government, involving different clinical populations. Consequently, and
because of its widespread use within the military, the
purpose of the current study was to extend the
available data on the SPRINT to include young adults
with normal hearing and older listeners with hearing
loss who were .60 years of age, which is above the age
range of the 319 soldiers on whom the SPRINT
originally was evaluated. A second purpose was to
examine the concurrent validity of the SPRINT by
comparing the recognition performances obtained on
the SPRINT by the younger and older listeners with
performances obtained on the Words-in-Noise test
(WIN) that also uses the NU-6 materials in a multitalker babble paradigm but at multiple S/Ns (Wilson,
2003; Wilson and Burks, 2005; Wilson and McArdle,
2007). The concurrent validity of the WIN was
established in previous studies (McArdle et al, 2005;
Wilson and McArdle, 2005; Wilson et al, 2007) that
examined the relationship between performances on
the WIN and other speech-recognition paradigms
including monosyllabic words in quiet, the HINT
(Hearing In Noise Test; Nilsson et al, 1994), the
BKB-SINTM (Niquette et al, 2003), and the QuickSINTM (Quick Speech-in-Noise test; Killion et al, 2004).
METHODS
Materials
The NU-6 words (Randomization C) used in the
SPRINT were recorded by Auditec of St. Louis (www.
auditec.com) (male speaker) and are mixed with a
Word Recognition in Noise Tests—SPRINT and WIN/Wilson and Cates
six-talker, multitalker babble at 9 dB S/N. Because 200
words were administered, the test required about 20 min.
The WIN consists of 70 NU-6 words spoken by the
VA female speaker (Department of Veterans Affairs,
2006). The WIN test has the following characteristics
(Wilson, 2003): (1) 70 monosyllabic words from the
same recorded version of NU-6 used for the recognition
measures in quiet (Department of Veterans Affairs),
which enables the evaluation of recognition performance in quiet and in multitalker babble with the
same materials spoken by the same speaker; (2) ten
unique words presented at each of seven signal-tobabble ratios in 4 dB decrements from 24 to 0 dB S/N;
(3) each word is time-locked to a unique segment of
multitalker babble, which reduces variability; (4) the
level of the babble, which is presented continuously, is
fixed and the level of the words varied; (5) the interval
between words is 2.7 sec; (6) the 50% point is
quantifiable with the Spearman-Kärber equation (Finney, 1952); and (7) a stopping rule that terminates the
test sequence when the ten words at one level are
incorrect. For the current study, the 70-word WIN list
was divided into two 35-word lists that were administered sequentially, and the results from the two runs
were combined (Wilson and Burks, 2005).
Half lists from List 4 of the NU-6 materials were
used to evaluate word recognition in quiet with the half
lists counterbalanced between two presentation levels.
The two levels, 60 and 80 dB HL, correspond to the
presentation level of the words in the 0 and 20 dB S/N
conditions of the WIN. Data from these two levels help
ensure that audibility was not an issue when interpreting the WIN data, especially at the less favorable
S/Ns. All of the speech stimuli were calibrated to a
common 1000 Hz tone using a vu meter (Tascam,
Model MU-40) and were recorded on a CD (Hewlett
Packard, Model DVD200i).
Subjects
Twenty-four young listeners between the ages of 19
years and 29 years (mean 5 23.3 years, SD 5 2.4) with
normal hearing and 48 older listeners between the
ages of 60 years and 82 years (mean 5 69.9 years, SD
5 6.7) with sensorineural hearing loss participated in
the study. The young listeners were recruited from the
local university and had normal hearing (#20 dB HL;
American National Standards Institute, 2004) at the
250–8000 Hz octave intervals. The inclusion criteria
for the listeners with hearing loss were as follows: (1)
age between 60 and 85 years, (2) a threshold at 500 Hz
between 15 and 30 dB HL, (3) a threshold at 1000 Hz
between 20 and 40 dB HL, (4) a three-frequency puretone average (500, 1000, and 2000 Hz) of #40 dB HL,
(5) word-recognition scores in quiet $40%, and (6) no
history of middle ear or retrocochlear pathology as
Figure 1. The mean audiogram for the right ears of the 48
listeners with hearing loss is shown along with the standard
deviations represented by the vertical lines.
determined by an audiologic evaluation. The pure-tone
criteria applied to the ear with the better thresholds.
Pure-tone symmetry was not considered, as wordrecognition performance in noise is essentially the
same (61 dB) for monaural and binaural conditions
when listening under earphones (Holma et al, 1997;
Wilson, 2003). The average difference between the
pure-tone thresholds for the right and left ears of the
listeners with hearing loss was ,1 dB. The average
audiogram and standard deviations for the right ear
are shown in Figure 1.
Procedures
Data collection occurred during a 40–60 minute
session. Following the informed-consent process,
pure-tone thresholds were obtained. The speech materials were presented bilaterally in the following order:
(1) SPRINT protocol administered at 50 dB HL, which
is the presentation level used by the Army, (2) two half
lists of NU-6 words were administered in quiet with
the first half list at 60 dB HL and the second half list at
80 dB HL, and (3) the two 35-word lists of the WIN
materials were administered with the multitalker
babble fixed at 60 dB HL. For the SPRINT the
instructions given in the Cord et al (1992) paper were
used. The speech materials were reproduced on a CD
player (Sony, Model CDP-497) that was routed through
an audiometer (GSI, Model 61). All stimuli were
presented under TDH-50P earphones while the
participants were seated in a double-walled sound
booth (IAC, Model 1204). With the word stimuli, the
551
Journal of the American Academy of Audiology/Volume 19, Number 7, 2008
Figure 2. The mean datum points obtained from the 24
listeners with normal hearing (circles) and the 48 listeners with
hearing loss (squares) for the SPRINT (large symbols with ‘‘S’’
included), for the WIN (small open symbols), and for the NU-6
presented in quiet (filled squares). The vertical lines through the
datum points represent 61 standard deviation. The lines through
the datum points are the best-fit, third-degree polynomials used
to describe the data.
verbal responses of the listeners were recorded into a
spreadsheet.
RESULTS AND DISCUSSION
T
he overall results from the 24 listeners with
normal hearing (circles) and from the 48 listeners
with sensorineural hearing loss (squares) are shown in
Figure 2. The two mean data points for the SPRINT
are depicted by the larger symbols denoted by ‘‘S’’ at
the 9 dB S/N. The mean WIN datum points are
depicted with the smaller symbols connected with
third-degree polynomials used to describe the data.
The vertical lines through the various datum points
indicate 61 standard deviation.
Consider first the SPRINT data in Figure 2. The
mean for the listeners with normal hearing was 92.5%
correct (SD 5 2.4%), whereas the mean for the older
listeners with hearing loss was 65.3% (SD 5 11.2%).
The 92.5% for the listeners with normal hearing was
slightly below the performance level targeted by Cord
et al (1992) of 95–100% with only five of the listeners
performing in the targeted range. Performance by the
listeners with hearing loss was on average 27.2% below
the mean performance by the listeners with normal
hearing and 16.2% below the mean performance
552
(81.5%) reported by Cord et al for the 319 soldiers
with H-3 profiles. This lower performance by the
listeners with hearing loss in this study was expected
as they were older than the Cord et al sample and they
had poorer pure-tone thresholds than the Cord et al
soldiers, especially in the 2000–3000 Hz range.
Although the SPRINT is scored as a composite of 200
words, the performances on each of the four NU-6 lists
were available. For the listeners with normal hearing
the scores for Lists 1-4 were 89.0, 93.7, 92.8, and 94.4%
with standard deviations ranging from 3.0 to 4.4%. The
respective scores for the listeners with hearing loss
were: 62.1, 66.8, 67.3, and 65.2% with standard
deviations ranging from 11.6 to 13.1%. Although no
statistical analyses were conducted on these data, it is
interesting that performance by both groups of listeners on List 1 was about 5% (,1 dB, assuming a slope of
5%/dB) poorer than the performances on the other
lists. Probably this relation is reflecting a ‘‘learning
curve or practice effect’’ by the listeners that occurred
during List 1, which was always given first. It is well
established that practice effects improve performance,
especially in difficult listening conditions such as noise
at less than optimal S/Ns (e.g., Wilson, Bell, et al, 2003;
Sweetow and Sabes, 2006). List 1 enabled the listeners
to improve their listening skills in the multitalker
babble background noise that was reflected by the
improved performances on the remaining lists. Alternatively, the difference in performances between List 1
and the remaining lists could be attributed to interlist
differences. This reason for the differences probably
can be discounted as the Auditec recordings of the NU6 materials have been shown to have good inter-list
equivalency (Wilson et al, 1976).
The WIN data in Figure 2 illustrate the typical
relation between recognition performances on the WIN
by listeners with normal hearing and listeners with
hearing loss. From the functions in the figure, the 50%
points were 1.8 and 11.3 dB S/N, respectively, for the
two groups of listeners, which is a 9.5% difference. The
slopes of the respective functions at the 50% points
were 8.1 and 6.7%/dB, again an expected relationship
with the function for the listeners with hearing loss
being more gradual owing to their increased variability
that is demonstrated by the standard deviations
(vertical lines through the datum points in the figure).
The filled squares in Figure 2 depict the recognition
performances in quiet by the listeners with hearing
loss at 60 and 80 dB HL, which correspond to the
presentation levels of the words in the 0 and 20 dB S/N
WIN conditions (the same data for the listeners with
normal hearing were .98% correct and are not plotted
in Figure 2). At 80 dB HL, the performances by the
listeners with hearing loss were essentially the same
in quiet (84.7%) and in multitalker babble (86.7%)
that indicates the babble provided no masking or
Word Recognition in Noise Tests—SPRINT and WIN/Wilson and Cates
Figure 3. Bivariate plot of the recognition performance on the
SPRINT (ordinate) and on the WIN (abscissa) for the 24 listeners
with normal hearing (circles) and the 48 listeners with hearing
loss (squares). The dashed lines represent the linear functions
used to describe the data, and the large filled symbols represent
the mean datum points. The shaded region of the graph defines
normal performance on the WIN (Wilson, Abrams, et al, 2003).
interference at the 20 dB S/N condition. In contrast,
when the presentation level was 60 dB HL, recognition
performance decreased in quiet slightly to 75.7%;
however, in babble at the same speech presentation
level, performance decreased dramatically to 0.2%.
Because performance in quiet was so relatively good at
60 dB HL, the lack of performance in babble can not be
attributed to the audibility factor but, rather, is
accounted for by the detrimental effects caused by
the multitalker babble. Likewise, at 60 dB HL,
audibility did not account for the poor performance
by the listeners with normal hearing whose performance dropped from 98% in quiet to 33.8% in babble.
Finally, from Figure 2, consider the relation between
the performances by the listeners with hearing loss on
the SPRINT and the WIN shown by the horizontal
separation between the large square and the WIN
function. On the SPRINT, 65% correct was achieved at
9.0 dB S/N, whereas on the WIN the 65% point was
achieved at 13.6 dB S/N (calculated from the polynomial equation), which is a 4.6 dB difference. Based on
data obtained in quiet in earlier investigations, this
4.6 dB difference is attributable to speaker differences
between the two recordings used by the SPRINT and
WIN paradigms. Wilson et al (1990, figure 3) analyzed
the psychometric functions generated by four recordings (different speakers) of the NU-6 lists in quiet,
including the Auditec recording (used with the
SPRINT) and the VA recording (used with the WIN).
Although the slopes of the functions were similar, the
displacement between the performances on the Auditec
and VA recordings of NU-6 in the Wilson et al study
differed by about 5 dB, with the VA recording
requiring a higher presentation level for equal performance. This is the same relation between the two
recorded versions of NU-6 that was observed in the
current study.
Figure 3 is a bivariate plot of the performances by
the individual participants on the SPRINT (ordinate)
and WIN (abscissa). The shaded region of the graph
defines normal performance on the WIN (#6 dB S/N;
Wilson, Abrams, et al, 2003). The WIN data are the
50% points derived with the Spearman-Kärber equation that produced means (larger, filled symbols) of 2.7
and 12.0 dB S/N for the listeners with normal hearing
and listeners with hearing loss, respectively (SDs 5 1.5
and 3.4 dB). With one exception, the listeners with
normal hearing are grouped closely above 90% correct
on the SPRINT and in the normal range of the WIN.
Because the scales on the ordinate and abscissa are
different, the apparent inverse relation between
performances on the SPRINT and WIN for the
listeners with hearing loss in reality indicates that as
performances improved on the SPRINT, performances
likewise improved on the WIN. Three of the listeners
with hearing loss had performances on the WIN that
were within the normal range, whereas one of the
listeners with normal hearing was outside the normal
range. A Pearson product-moment correlation indicated a significant relationship (r 5 20.81, p , .01)
between the performances on the SPRINT and WIN.
This highly related correlation indicates good concurrent validity of the SPRINT (Franzblau, 1958), which
was one of the objectives of this project.
Pearson product-moment correlations indicated significant relationships between performances on the
NU-6 materials presented at 80 dB HL and on the
SPRINT (r 5 0.71, p , .01) and on the WIN (r 5 20.75,
p , .01). This marked relation for the SPRINT data is
illustrated in Figure 4, which is a bivariate plot of the
performance on the SPRINT (ordinate) versus performance on NU-6 presented at 80 dB HL in quiet
(abscissa) by the listeners with normal hearing (circles)
and by the listeners with hearing loss (squares). The
large, filled symbols depict the mean data, and the
dashed line is the linear regression used to describe the
data. The data from the listeners with normal hearing
are near maximum for both conditions. In general, for
the listeners with hearing loss there is a direct relation
between performances on the SPRINT and NU-6 in
quiet, but the average difference between the two
measures was 19.4%. Although several listeners with
hearing loss had 90–100% scores in quiet, most of the
SPRINT scores were ,80%.
553
Journal of the American Academy of Audiology/Volume 19, Number 7, 2008
Figure 4. A bivariate plot of the recognition performance on the
SPRINT (ordinate) and on the NU-6 in quiet (abscissa) by the 24
listeners with normal hearing (circles) and the 48 listeners with
hearing loss (squares). The dashed lines represent the linear
functions used to describe the data, and the large filled symbols
represent the mean datum points.
The relations between the traditional pure-tone
average (PTA at 500, 1000, and 2000 Hz) and performance on the SPRINT and between the high-frequency,
pure-tone average (HFPTA at 1000, 2000, and 4000 Hz)
and performance on the SPRINT are depicted in
Figure 5. Two relations are apparent from the data in
the figure. First, as expected from the audiogram in
Figure 1, the distribution of datum points and the mean
for the HFPTA variable are displaced to the higher
hearing levels than are the datum points and mean for
the traditional PTA. Second, the slope of the linear
regression used to describe the data for the HFPTA
variable is steeper than the slope of the function for the
traditional PTA. A steeper slope indicates a better
relation between the two variables under analysis. The
second point is supported by the Pearson productmoment correlations that for the SPRINT were significant, but at different levels, for both pure-tone averages.
The SPRINT and the traditional PTA had a fair relation
(r 5 20.31, p , .05), whereas the SPRINT and the
HFPTA had a moderate relation (r 5 20.52, p , .01).
Although not illustrated, the relations between the WIN
data and the two pure-tone averages were different from
the relations observed with the SPRINT. There was not a
significant relation between the WIN and the traditional
PTA, whereas there was a moderate relation (r 5 0.56, p
, .01) between the WIN and the HFPTA. This was an
expected result, as a number of earlier studies have
observed this differential relation between word recog-
554
Figure 5. Bivariate plots of performance on the SPRINT
(ordinate) by the 48 listeners with hearing loss and the average
pure-tones (abscissa) for 500, 1000, and 2000 Hz (top panel) and
1000, 2000, and 4000 Hz (bottom panel). Overlapping datum
points were jittered to avoid overlapping datum points. The
dashed lines are the linear regressions used to describe the data,
and the larger, solid symbols represent the mean data.
nition in noise and the two pure-tone averages (e.g.,
Kryter et al, 1962; Harris, 1965; Wilson and McArdle,
2005, figure 5). Interestingly, the traditional PTA was
not significantly related to understanding speech in
quiet, whereas the HFPTA and understanding speech in
quiet were moderately related (r 5 20.43, p , .01).
Figure 6 contains histograms of the number of the
200 SPRINT words whose percent correct was in each
of ten percent-correct intervals. The histograms for the
listeners with normal hearing (top panel) and with
hearing loss (bottom panel) are shown. As overall
performance on the SPRINT by the listeners with
normal hearing was 92.5% correct, the majority of
words that were correct were in the 90–100% interval.
The distribution of correct words with the listeners
with hearing loss was quite different with an almost
flat distribution across the percent correct intervals
from 30–100%. The maximum number of words in any
interval was 40 in the 80–90% interval. This unimodal
Word Recognition in Noise Tests—SPRINT and WIN/Wilson and Cates
3. Graphically and numerically the SPRINT and WIN
were highly related, which is indicative of good
concurrent validity of the SPRINT.
4. Recognition performances on the NU-6 in quiet and
the SPRINT and on the NU-6 in quiet and the WIN
were correlated significantly.
5. The HFPTA (1000, 2000, and 4000 Hz) had higher
correlations with recognition performance on the
SPRINT, WIN, and NU-6 in quiet than did the
traditional PTA (500, 1000, and 2000 Hz).
Acknowledgments. The Rehabilitation Research and Development Service, Department of Veterans Affairs supported this work through a Merit Review, the Auditory and
Vestibular Dysfunction Research Enhancement Award Program (REAP), and a Senior Research Career Scientist award
to the first author. The authors acknowledge the contributions made by to the project by Kelly Koder, Rachel McArdle,
Sherri Smith, and Brian Walden.
REFERENCES
American National Standards Institute (ANSI). (2004) Specification for Audiometers (ANSI S3.6 2004). New York:
ANSI.
Cord MT, Walden BE, Atack RM. (1992) Speech Recognition In
Noise Test (SPRINT). Walter Reed Army Medical Center.
Figure 6. Histograms of the number of SPRINT words identified correctly for the various percent correct categories determined with the 24 listeners with normal hearing (top panel) and
the 48 listeners with hearing loss (bottom panel).
distribution of words correct at a given presentation
level helps explain the intersubject variability that is
associated with word-recognition scores. No doubt, had
the data from the listeners with normal hearing been
obtained at a less favorable S/N at which recognition
performance was 40–60% correct, a similar distribution of words correct would have been observed.
SUMMARY AND CONCLUSIONS
T
he results from the current study are summarized
as follows:
1. Recognition performance on the SPRINT by young
listeners with normal hearing (92.5%) was close to the
95–100% performance by listeners with normal hearing targeted by Cord et al (1992).
2. The older listeners with sensorineural hearing loss
included in this study performed poorer on the
SPRINT (65.3%) than did the younger group of Army
personnel with H-3 profiles (81.5%) reported by Cord
et al.
Department of the Army. (2007) Standards of Medical Fitness. Army Regulation 40-501. Washington, DC: Department of the Army. http://www.army.mil/usapa/epubs/pdf/r40_
501.pdf.
Department of Veterans Affairs. (1952) Speech Recognition and
Identification Materials. Disc 4.0. Mountain Home, TN: VA
Medical Center.
Finney DJ. (1952) Statistical Method in Biological Assay London:
C. Griffen.
Franzblau AN. (1958) A Primer of Statistics for Non-statisticians.
New York: Harcourt, Brace and World.
Harris JD. (1965) Pure-tone acuity and the intelligibility of
everyday speech. J Acoust Soc Am 37:824–830.
Holma T, Laitakari K, Sorri M, Winblad I. (1997). New speech-innoise test in different types of hearing impairment. Acta
Otolaryngol Suppl 529:71–73.
Killion MC, Niquette PA, Gudmundsen GI, Revit LJ, Banerjee S.
(2004) Development of a quick speech-in-noise test for measuring
signal-to-noise ratio loss in normal-hearing and hearing-impaired
listeners. J Acoust Soc Am 116:2395–2405.
Kryter KD, Williams C, Green DM. (1962) Auditory acuity and
the perception of speech. J Acoust Soc Am 34:1217–1223.
McArdle RA, Wilson RH, Burks CA. (2005) Speech Recognition in
Multitalker Babble Using Digits, Words, and Sentences. J Am
Acad Audiol 16:726–739.
Nilsson M, Soli S, Sullivan J. (1994) Development of the Hearing
in Noise Test for the measurement of speech reception thresholds
in quiet and in noise. J Acoust Soc Am 95:1085–1099.
555
Journal of the American Academy of Audiology/Volume 19, Number 7, 2008
Niquette P, Arcaroli J, Revit L, Parkinson A, Staller S, Skinner
M, Killion M. (2003) Development of the BKB-SIN Test. Paper
presented at the Annual Meeting of the American Auditory
Society, Scottsdale. http://www.amauditorysoc.org/pages/abstracts/
2003/2003-pp-1.html.
Sweetow RW, Sabes JH. (2006) The need for the development
of an adaptive Listening And Communication Enhancement
(LACETM) Program. J Am Acad Audiol 17:538–558.
Thornton AR, Raffin MJM. (1978) Speech-discrimination scores
modeled as a binomial variable. J Speech Hear Res 21:507–518.
Tillman TW, Carhart R. (1966) An Expanded Test for Speech
Discrimination Utilizing CNC Monosyllabic Words. Northwestern University Auditory Test No. 6. Technical Report. Brooks Air
Force Base, TX: USAF School of Aerospace Medicine.
Wilson RH. (2003) Development of a speech in multitalker babble
paradigm to assess word-recognition performance. J Am Acad
Audiol 14:453–470.
Wilson RH, Abrams HB, Pillion AL. (2003) A word-recognition
task in multitalker babble using a descending presentation mode
from 24 dB S/B to 0 dB S/B. J Rehabil Res Dev 40:321–328. http://
www.vard.org/jour/03/40/4/pdf/Wilson-B.pdf.
Wilson RH, Bell TS, Koslowski JA. (2003) Learning effects
associated with repeated word-recognition measures using
556
sentence materials. J Rehabil Res Dev 40:329–336. http://www.
vard.org/jour/03/40/4/pdf/Wilson-A.pdf.
Wilson RH, Burks CA. (2005) The Use of 35 Words to Evaluate
Hearing Loss in Terms of Signal-to-Babble Ratio: a Clinic
Protocol. J Rehabil Res Dev 42:839–852. http://www.rehab.
research.va.gov/jour/05/42/6/pdf/wilson.pdf.
Wilson RH, Coley KE, Haenel JH, Browning KM. (1976)
Northwestern University Auditory Test No. 6: Normative and
Comparative Intelligibility Functions. J Am Audiol Soc 1:221–
228.
Wilson RH, McArdle R. (2005) Speech signals used to evaluate
auditory function. J Rehabil Res Dev 42(Suppl. 2):79–94.
Wilson RH, McArdle R. (2007) Intra- and inter-session test, retest
reliability of the Words-in-Noise (WIN) Test. J Am Acad Audiol
18:819–831.
Wilson RH, McArdle RA, Smith SL. (2007) An evaluation of the
BKB-SIN, HINT, QuickSIN, and WIN materials on listeners with
normal hearing and listeners with hearing loss. J Speech Lang
Hear Res 50:844–856.
Wilson RH, Zizz CA, Shanks JE, Causey JD. (1990) Normative
data in quiet, broadband noise, and competing message for
Northwestern University Auditory Test No. 6 by a female
speaker. J Speech Hear Disord 55:771–778.